With the advancement of sensors for recording data having 3D information such as laser scanning, a stereoscopic camera, a structured light scanner, the Microsoft Kinect™, a light field camera, and the like, various types of 3D stereoscopic image data may be obtained. The respective sensors generate 3D data in different formats, and such data in different formats have the merits and demerits, which, thus, are stored in different manners and different algorithms are applied thereto to be processed according to purposes.
A framework of software for 3D stereoscopic image processing performs 3D stereoscopic image processing using various types of data formats. Since data formats are different, when the same stereoscopic image processing algorithm is applied, data formats are converted into intermediate data in a common format and the stereoscopic image processing algorithm is applied thereto. Thus, obtained 3D stereoscopic image data is converted into intermediate formats and the 3D image processing algorithm is subsequently applied to the converted intermediate format, not to the original obtained format.
However, there may be a native algorithm that may be used only in a data format, not a common algorithm, according to types of 3D image processing scheme and types of data. In this case, the native algorithm applied only to the corresponding format should be used to obtain better results in terms of speed and quality.
Also, data storing 3D information since it was obtained or generated (for example, a laser-scanned point cloud, 3D mesh generated through modeling by a user, and the like) has a significant difference in data characteristics from 3D image data based on a 2D image, so conversion of such data into a common format and application of a common algorithm thereto causes huge data loss.